Dynamic maintenance of decision rules for decision attribute values’ changing

Abstract

Rule induction method based on rough set theory (RST) which can generate a minimal set of decision rules by using attribute reduction and approximations has received much attention recently. In real-life, the variation of objects, attributes and attributes’ values affects reducts and approximations, e.g., the coarsening and refining of attributes’ values. The goal of this paper is dynamic maintenance of decision rules for decision attribute values’ coarsening and refining. Two incremental rough-set based methods are proposed to deal with this issue by updating assignment discernibility matrix dynamically without recomputing the reducts from the beginning, which increases the efficiency.